Re: Possible planner bug/regression introduced in 8.2.5

Hi Tom,

>> I tried to simplify the query even more and now I have query which on
>> 8.2.4 is planned instantly and on 8.2.5 takes cca 8 seconds.
>
> Are you sure you were using the same planner parameters (particularly
> join_collapse_limit and the geqo threshold) in both cases?

thanks for the reply.

Yes, I hope I'm using same parameters. To exhibit the behaviour I set
geqo to off and join_collapse_limit (and from_collapse_limit) to 1000.

On the test server, I can start postgresql 8.2.4 and 8.2.5 on the same
datafiles and there is still this big difference in planning time. I
hope I'm not missing something obvious and not wasting your time...

What can I do to help to debug it?

Thanks,

Kuba